A review of plastic-encapsulated-microcircuit reliability-prediction models

Abstract

This paper reviews available plastic encapsulated microcircuit (PEM) reliability prediction models including the CNET, Thomson-CSF and RAC model. We examine the RAC (Reliability Analysis Center) model in detail, as it is the only usable model that considers the accelerating effects of temperature and humidity on PEM failures. Our intention is to provide constructive criticisms and propose amendments. We agreed with the assumptions made by the RAC model: types of packaging do not affect PEM failure rate; and a single activation energy of 0.8 eV may be adequate for different IC logic families. However, we feel that the RAC model should incorporate a term for EOS/ESD failures as they constitutes a high percentage (up to 43%) of IC failures. The RAC model should set an upper limit of 130/spl deg/C and 99% RH when using Peck's model to model temperature/humidity acceleration so as to avoid violating physics-of-failure considerations. It should consider treating microprocessor and memories as two different classes of devices due to the vast differences in their reliability as indicated by some literatures. We question if the reliability growth rate can be projected without restrictions. There should also be clearer indications of what constitutes "best commercial practices"-a prerequisite to apply the model.